ChuckMcM

This is a much better introduction to circuits than most. You actually get a bit of depth rather than a cookbook of "do this to get that" sort of prose.

You realize that the slicer has to talk to the firmware, and it is documenting the commands their firmware implements which makes the combination of slicer + firmware print right? The whole point of the AGPL on the slicer was to prevent people from having special sauce printer firmware, get all the work of the slicer for "free" and then make their combination better than everyone else's. So you publish the code the slicer uses to talk to your device and that lets someone else implement the same codes (if they choose too).

From the article: With the LLMs every time there’s a mistake, I don’t learn. I can spot the mistake because I have some experience (lucky enough to have been born 34 years ago), I can pinpoint the mistake and let the LLM fix it.

This is something I realized when I became a manager of programmers rather than doing the work myself. When I write code I do it from a position of having a deep understanding of what all the moving pieces are, when someone working for me writes code I just have to trust they had that level of understanding[1].

But being a manger, the thrill of solving the puzzle was replaced by the thrill of helping some new engineer get better at the craft of writing code. They took pride in their growing skill set and I took pride in helping them grow.

Managing consultants however was very transactional. Did they deliver what was asked of them? Did they fix the flaws I pointed out? Do I still want to pay them?[2] And I could of course fire them easily and try new ones that had a reputation for being "better" at the tasks I needed done.

If history is any judge, there have been many companies over the years where some founder/leader has a "great idea" and just needed someone technical to make it real. And the typical arc for those companies was they get founded, they spend a ton of money on consultants to build their technology, something changes (fashion, systems, something) and their product is no longer a good fit for the market so they try to hire consultants to fix it, which now costs more than 2x because first the consultants had to understand what the previous consultants did, then they have to change it in a way that leadership wants. The result is a product that got more and more brittle until there wasn't enough money to or enough consultants to anneal the product back into something people would pay enough for to support the salaries of the visionary leaders who never had any idea at all of how the product they sold actually worked.

It feels like we're going to speed run that. In the LLM era.

[1] And of course that level of trust is not always warranted but that's a different comment :-)

[2] With consultants there is a velocity vs organizational expertise equation that always needs to be balanced. Dropping people into a legacy code base they've never seen is just as hard for them as dropping them into a consultant's work product that they hadn't previously seen.

I think the author meant 'burnish' there, which is a clever way of showing they didn't use AI :-)

I appreciate where you're coming from, but I think you may have unintentionally made my point.

Try this: Change the context "Acme's Sunscreen which is made from killing an endangered fish, is significantly better than any other sunscreen out there, and because it is already on fish we know it doesn't hurt the ocean!" except the production of Acme Sunscreen is leading to the extinction of said fish in the future. Short term profits, long term loss, and perhaps a bet that at some point what ever it is about the fish that makes great sunscreen will be figured out.

Using FOSS requires (by license) that you give back your changes, which, when functional prevents a vendor from exploiting the work of a bunch of people they didn't pay, to make a product where they pay a few people to push it way past what the community offers. Whereas if they had been publishing their changes, then their product wouldn't be "significantly ahead" because everyone using the same slicer base code would be able to also use Bambu's improvements. But Bambu doesn't do that because why let the competitors take advantage of the work of people who they didn't compensate for that work. Hmmm?

See how that works?

Yes. Well they certainly bill at that rate, it goes some to the lawyer, some to rent, etc.

Inevitably the situation is the "sketch" here is that the vendor avoids a cost by creatively interpreting the requirements. It became famously difficult to buy a reliable PC power supply at one point because PSU vendors were all trying to cut costs and undercut each other. When there isn't a lot of market differentiation features, people differentiate on price. At NetApp we got more and more prescriptive in our contracts with motherboard OEMs because they would do silly things like "leave off 1/2 the bypass capacitors" which would then cause the filer to fail in some environments. And I get that "its just business"

It's an easy thing to avoid the costs associated with complying with FOSS licenses because they FOSS people don't come after you. So instead of having a person on staff who is making sure the company is compliant and updating the web site with source as it changes etc etc. You just don't do any of that and not only do you save money not paying that person you don't have to bother the web site people every time you do a software update. So from a manufacturer's perspective it's a win.

That said, Chinese vendors selling to Chinese customers following Chinese law. All good. Chinese vendors selling to US customers and not following US law, that's bad. I don't see how Chinese law would be part of that discussion.

It was really depressing when I realized that people will buy a sketchy product which is hurting the community and makers of non-sketchy products because "its cheaper and its not like my purchase is gonna change things one way or another."

It is selfish and short sighted, and it leaves their future impoverished, but in the moment they "got a good deal."

For background, I participated in a "small" action that Sun took against someone who was violating the Java license. The legal fees (in the late 90's) were 8.2 million dollars. $250K is nothing to sneeze at, its a solid 6 weeks of a single lawyer's time, maybe a couple of weeks for a small team. Against a company doing nearly a billion dollars a year in revenue[1] and owning 29% of the printer market? Add in a couple of expert witnesses which are essential to that sort of a case and you'll see that $250K isn't even table stakes. Sad, I know, but it is the way things are these days.

And the saddest thing is that Bambu knows this. They know that people who give away software are not a serious threat to them. And it is especially galling that the value they have extracted out of that software will pay the lawyers that keep you from preventing them from continuing to extract that value and keep it for themselves.

[1] "Media reports have also indicated that Bambu Lab has shipped more than 1.2 million units, claiming a 29% global share." -- https://kr-asia.com/bambu-lab-nears-usd-100-billion-valuatio...

I think the Bambu case is a good one to litigate the AGPL but perhaps should start in the Court of International Trade. That court has the power to block imports as part of a TRO which would shut down a lot of their revenue from the US at least. Doing something similar in Europe would also help apply pressure. Takes lawyers and money though, of which there is certainly not enough of the latter.

Not to be too glib here, but its one of the paradoxes of "free software" which is there isn't any money generated by the licenses that keep it free. Where as "licensed" software has a revenue stream that is used to enforce the license against people who fail to honor it. The FSF can only do so much with their budget here.

I think 'least regard' belongs to Oracle. For some reason I thought DuckDB was one of Gabe's ideas over at DuckDuckGo :-).

That said, the acquisition by Amazon is interesting in that it seems to overlap with their existing DB offerings? Which makes me wonder what exactly it is that they are getting out of this.

It is pretty amazing the creativity with compute architecture which is afforded by a zillion transistors.

Its a fun radio, the Si4844-A10 is another one and it can tune AM as well. I built one for my house that was connected to an STM32 that was connected to the Internet and would let me stream AM radio to my phone. That way I could listen to sports when I was on the road with the "home" announcers :-).

The Silicon Labs folks did a write up on their radio with an Arduino here: https://community.silabs.com/s/share/a5U1M000000knsEUAQ/buil...

> RISC-V is .. fine.

Yeah, so was 8051 and it sucked too :-). I appreciated having this rant all in one place. Ranting against bad architecture is always cathartic and absolutely useless since the people who built and now champion the bad architecture are invested so one's rant simply irritates them. And like the parent comment here, I too find RISC-V "useful" in that it has sufficient tooling to make most everything foundational 'out of the box' rather than me having to build it.

Perhaps the most interesting thing is that RISC-V shows just how ISA agnostic people are, as long as you have cross compilation with the gcc suite and an open source way to program and debug things. Before RISC-V, working on a bespoke ISA and computer architecture was never going to "go" anywhere except perhaps into a paper or conference talk. Now there is evidence of a non-zero chance of it going mainstream. :-)

Which sort of affirms the mantra, "If you can't prove causation, try to prove correlation." This because you can market correlation to unsophisticated readers and imply causation.

> The things that make me me are highly deterministic!

Are they though? :-) There are some interesting papers in the tissue regeneration space which are working on building tissue (and organs) from stem cells for medical purposes (transplants, injury treatment, Etc.) and one of the things that comes out from that is that a set of stem cells make unique tissue every time in that it's compatible but the fine structure is always randomly different!

While the growth of brain matter is a minefield of ethical issues, at some point I suspect we're going to have to figure out how to do that to treat things like TBI and neurodegenerative diseases. In terms of understanding how randomness plays a part in your existence though cellular biology papers are a pretty good source.

Great writeup, and it was great to see them step in an pay the developers of SQLite to help them fix the bug. I get tired of corporations asking open source authors to fix problems that affect the corporation for free. And while I'm sure it was frustrating for folks to have these outages, I find such puzzles pretty fun to get to the bottom of.

I completely agree, tremendous complexity can arise from simple mechanisms. Gleick's Chaos is a really good introduction to that. I was talking about the article though, and the mechanisms currently used for training and inference in LLMs. Those mechanisms are mathematically precise (unlike Chaotic attractors) and as the author points out, achieve the same function as compressors do in a strict bit pattern minimization role. Sometimes tensor math is pretty complex, like the FFT and DCTs on JPEG compression, but with the same inputs you get the same results. And while a JPEG will never decompress to a different image than the one that was compressed in the first place, LLMs do not 'infer' token streams that haven't been trained in their training process. The big difference here is that if you imagine a JPEG compressor that compresses 100 different images into one 'chunk', you can see how to provoke it to produce any one of the images it previously compressed. And with a bit of creativity you can have it express different images in different parts of the resulting composite. FWIW I looked at patenting something like this for digital cameras to give them more "shots" space for a given amount of SD storage.[1]

Given the way that models work in 'inference' mode (vs 'training' mode) you can't forward bias the result into the correct result when there are multiple forward results that have identical weights. It's the root cause of hallucinations, and you've lost information in the training phase that you can't then use to discriminate between the 'right' answer and an equally valid 'wrong' answer.

[1] FWIW I could never recover enough state to insure that the image it regenerated was all of the same image you took. So you might get the street but one of the houses might be a house that was in a different picture you took. That kind of bug. Mostly arising out of the same kind of problem you have with using hashes to find documents, when you get a hash collision two documents have the same hash, so you don't know which one to return.

I suspect you're being snarky :-) but this is a really interesting question, and one that has had a lot of research done. I'm not current (I stopped following folks doing this research closely around 2019) but what we 'didn't' know about how brains work was still huge. Signaling levels, enzymes, the connectome, quantum effects, it is a really deep question.

That said, once we do get a working idea of how it works, and can perhaps synthesize a brain artificially with proteins, it will inform us on the next steps for silicon realization of that.

It also helps explain to people that LLMs are as likely as bzip to develop "consciousness".

This is, in general, a good statement of a durable problem one can 'solve' profitably. Basically take a problem that is hard to do 1:1, systemize it such that you can easily tune the solution to "all" variants of that problem, and then sell that as a service taking a percentage which is still going to be less than the cost of the customer doing a one-off solution.

From the article -- In coffee shops and in the low-lit corners of bars, professionals are sharing dreams of “disappearing” or “living on a farm somewhere” or “going off the grid.”

Because I spent my career working in tech in the Bay Area there are some things that I just think of as 'normal' that other people, without that exposure, are surprised by. During the last 40 years start ups and tech companies have minted tens of thousands of millionaires across a wide distribution of 'net worth just over a million dollars' to 'net worth of tens of millions'. The largest chunk of that cohort leave tech directly and go do 'fun' work. The smallest chunk of that cohort are still working for someone else trying to get a bigger score. What I realized was the most people aren't sociopathic wannabe billionaires. They don't get headlines, they don't show up "out of place", they are just the people who are living their lives doing things that matter to them.

So when the author is on about 'sharing the dream of doing what you want' I was like, "Yup, same as it ever was." The author also does not appear to have the benefit of having been in the area when semiconductor manufacturing became 'foundries' and all the process engineers and other workers in all of the small fabs in the SF Bay Area were now displaced. That was a class of workers who lost faith in their careers. But what was true then, and is true today, in the US we take kids out of High School and spend four years teaching them how to think and generally how to think about some topic in college and then they go out and get jobs. For someone who has been through college, getting a new Bachelor's degree in an entirely new subject takes two and a half years tops. You don't need the general ed, if its another STEM topic you don't need to redo the math, its just "lets take the classes that are specific to the <foo> major."

And before everyone jumps on this and says "But Chuck, now I've got a mortgage and responsibilities and I can't be taking a couple of years off just to go to school." Yeah I get it. That's especially true if you don't have a support network. So it sucks and is depressing and you don't want to even get out of bed because you feel like years of your life were just wasted and now you need to start over. The alternative though just gets you deeper and deeper into a hole.

In my opinion it carries much truth. Somewhere in Google's archives there is a message I sent to the OC that said, "The department of philosophy does not ship solutions to real problems, that is why there has never been a successful company that insisted the world adapt to its philosophy." Google was, even then, blind to what they were missing.

In my original comment I linked the nature study that looked at farmers doing both. The "but if we don't plant crops we'll starve!" argument is just silly for a number of reasons, one I pointed out which is if we used all land arable land just in California for solar panels + batteries we'd have 1,000 times as much power available as the entire world demand today. Who would even buy that power? Nobody. The thing that keeps this from happening is that as more solar power comes online its cost goes down and it becomes uneconomical to add more (farmer's won't put it in because they can make more money growing beans or something). So no, allowing farmers to put in solar instead of crops will never lead to a famine due to nobody growing food. People are too greedy :-). There is a large ConAgra farm in the central valley that lobbies a lot (they want more water for Almonds because $$) and a back of the envelope calculation shows they could convert half their acreage to solar, use that to desalinate water off the coast of Point Magu, and pump it to the other half of their acreage and not need any water delivered from the California watershed. It sounds crazy but its the new reality of what you can do with solar and batteries. China (as a government policy directive) has figured this out before anyone else btw.

The "Elevator Problem" (and the "Street Light Problem") were both CS 200 exercises when I was in college. Along with a bunch of other 'game' type systems like 8 queens, Hanoi towers, maze solving, etc. Still, that is the most elegant web widget I've ever seen that illustrates the options. If you're reading this John I'd love to hear how you wrote them.

Totally going to reach out here and see if we can help each other.

Yes, if farms went 100% from agriculture to solar we wouldn't have enough food :-) And we'd have anywhere from 100x to 1000x the entire power demand of the planet in energy. Solar has a $/acre value (which at the moment is pretty high in California) but as more Solar came on line that would go down. And like any crop they would figure out what they needed to maximize profits.

It might also spur the development of 100% electric Farm vehicles (tractors, planters, combines etc) as the Farm could make it own electricity and not have to pay diesel costs for fueling that. At the extreme, if you have enough 'free power' on your farm you can use it for electrolysis of Ammonia to make nitrates. (fertilizer). That would be because you could make money offsetting the other costs of farming to make your cost of goods lower than your neighbors farm and increase your total $/acre.

So the short summary is "no" is isn't going to change how much food is available. Farmers pick crops all the time that aren't "food" for you, they are exports or added to some other product (corn syrup, cotton), because they are in demand. That balances out by regular market dynamics.

Its a fair question, California agriculture tends to use 11 trillion gallons (34e6 acre feet[1][2]) of water annually. (about 1,700 Googles worth). Generally from a market driven standpoint agricultural users are trying to get maximum $/acre out of their land. I'm trying to convince the state legislature to reclassify "solar farms"[3] as agriculture (as opposed to 'industrial' that they are today) which would allow a farm to decide on an acre by acre basis how much to plant and how much to put into solar farming. Presuming batteries were included in that allowance, 350,000 acres of California farmland as solar farms (panels + batteries for 24/7 power) would power the entire state 24/7 with no need for coal, gas, or nuclear power plants. 100% green energy.

[1] https://en.wikipedia.org/wiki/Water_in_California

[2] https://www.convertunits.com/from/billion+gallons/to/acre+fo...

[3] "Agrivoltaics as a climate-smart and resilient solution for midday depression in photosynthesis in dryland regions" -- https://www.nature.com/articles/s44264-025-00073-1

In the US at least there is a lot (and by that I mean like maybe more than half) of civil case law around seller liability for defective or 'dual use' products. In the 70's some cities tried to sue hardware stores for selling spray paint that taggers were using, in several jurisdictions you can find authorities trying to sue vendors of lock picking and/or safe opening tools, etc. My non-lawyer reading of all that is that if it is reasonable to assume that the vendor didn't know, at the time of sale, what the customer was going to do with it, they aren't liable.

Once a vendor has been notified that these units are doing these sorts of things they will stop selling them. Its sadly very prescriptive in that if Newegg gets a notice that "WatchFunTV" streaming sticks are doing this, they will remove that brand but if the same hardware shows up from the same vendor as "SuperTVStreamer" or some such, that product won't be banned until someone does the test and then notifies the sellers. It's cat and mouse all the time.

Now the people who could do something about it, the ad networks like Google, do not do anything because ad revenue is ad revenue, people buying the ads cannot prove that the click was false so hey who can say it was? Which is why ad fraud is a perennial favorite of crooks. The people being ripped off don't have any way to prove it without a lot of support from the ad network traffic data which is "proprietary". Really stupid ad fraud gets shut down, but put a bit of care into it so that the Ad network and claim ignorance? You can do that all day. Just don't get greedy and try to pull in more than say 30 or 50 thousand dollars a month. Remember, the IAB said in 2025 alone Ad Revenue was $300B[1] so 2% of that is only $6B and any network with 2% or less of undetected fraud is considered a "high quality" ad network.

So yeah, ad fraud is the gift that keeps on giving.

[1] https://www.iab.com/insights/internet-advertising-revenue-re...

Yeah pretty much. Was the future though, amirite?